← Research Library
New to these reports? Start here
  • Dotted-underlined words have a plain-English definition — hover or tap them. Every term is also on the glossary page.
  • "Null" means we found nothing, not that something broke. Most reports here are negative results, on purpose — knowing an idea doesn't work is the point.
  • Two questions get asked separately. First, is the effect real? Second, is it already priced into the betting odds? An effect can be completely real and still useless to bet on.
  • A "calibration" row is a self-check. It runs the same method on something already known to be true. If that fails, the whole report is unreliable — so it's reported alongside the findings.
  • If a confidence interval includes zero, the real effect might be nothing at all, so no claim gets made.

Is Kyler Murray Worse After a New Call of Duty Release?

One-off study, 2026-10-21 (written up 2026-10-27). Prompted by the running joke that Kyler Murray, a known Call of Duty player, falls off when each year's new game comes out in late October or November.

Verdict: nullnull resultA test that found nothing. "Null" is the starting assumption that there is no real effect; a "null result" means the data gave us no reason to abandon that assumption. It does not mean the data was missing or the test failed to run. for release weeks; a weak lean for Double XP weekends (below). After adjusting for the time of year, Kyler is no worse after a Call of Duty launch than any other QB. Only 4 seasons could be tested, so even a real effect would be too small a sample to show.

Method

  • Releases: the US release date of each year's Call of Duty title, 2019-2025.
  • Measure: EPAexpected points addedHow much a play changed the number of points its team should expect to score on that drive. A better measure of a play than raw yards. per dropback, regular season, from nflverse play-by-play.
  • Windows: the 3 weeks after release (days 0-20) vs the 4 weeks before (days -28 to -1). A QB-season counts only with at least 40 dropbacks in both windows.
  • Comparison: the same before/after change for every other QB-season (187). Releases land in late October to mid-November, when passing efficiency dips league-wide (weather, injuries, defenses settling), so Kyler's change has to be judged against everyone else's.

Results

Season Game Before After Change
2019 Modern Warfare +0.21 +0.01 −0.20
2020 Black Ops Cold War +0.27 +0.12 −0.16
2022 Modern Warfare II −0.01 −0.06 −0.06
2024 Black Ops 6 +0.09 +0.41 +0.32
  • Kyler's average change: −0.025 EPA per dropback. Other QBs over the same weeks: −0.031 (sd 0.23, n = 187). Everyone dips a little.
  • Half of random 4-season draws from other QBs look as bad or worse (50.8% of 20,000 draws). No signal.
  • 2019 and 2020 were real drops, and are probably where the joke started. 2024 went hard the other way: after Black Ops 6 he had his best stretch of the year, including +0.49 at Miami two days after release and +0.72 in the 31-6 win over the Jets.

Seasons that couldn't be tested

Season Game Why
2021 Vanguard Hurt his ankle a week before release; no "after" window
2023 Modern Warfare III Returned from the ACL tear two days after release; no "before" window
2025 Black Ops 7 On injured reserve before release
2026 — Not tested; no release date at the time of the study

Caveats

  • Four seasons is far too few to rule out a small effect; this only shows there is no large or consistent one.
  • Release date is a stand-in for "playing a lot of Call of Duty"; nothing here measures how much he actually played.

Double XP weekends (added 2026-10-27)

Verdict: a lean, not evidence. Kyler is a bit worse in games played during a Double XP event, but how much depends on which events you count, and the one cut that looks significant is small and was one of several tried.

Method (double_xp.py). Each QB-game's EPA per dropback is adjusted for that QB's own season average and for the league-wide level that week. Kyler's adjusted value in games inside a Double XP window is compared with other QBs' games on the same dates (20,000 draws, one same-day game per Kyler game). Regular season and playoffs 2019-2025, games with at least 15 dropbacks.

Events counted Kyler games Kyler (adjusted) Other QBs, same dates Random draws as bad or worse
All 28 listed 15 −0.056 +0.019 14%
Confirmed only (13) 8 −0.157 +0.017 3%

Raw, before adjusting: +0.04 EPA per dropback in Double XP games vs +0.11 in his other games (all events); −0.09 vs +0.11 (confirmed only).

Why it's only a lean: - It depends on the cut. The 7 games in single-source windows average +0.06, so the full list shows nothing. There's no reason the real effect would only show up in better-documented events. - Several tests were run (release windows, all events, confirmed events), and the 3% is one-sided before allowing for that. - Eight games. Dropping the 34-7 loss to the Rams on 2019-12-01 (−0.58) takes the confirmed mean to −0.10; 6 of the 8 games are below his level. - The calendar is incomplete. No archive lists every event; dates were pieced together from news coverage. September and October events before 2025 are mostly missing. An unlisted event waters the test down, for every QB equally.

The fair test is forward: log the 2026+ Double XP windows when they're announced and check his games in them, without changing the method.

Double XP calendar sources

Files

  • kyler_cod.py: the full study, rerunnable (needs the per-season play parquet the stats pages use).
  • results.json: per-season windows, comparison group, and Kyler's games within 35 days of each release.
  • double_xp.py / double_xp_results.json: the Double XP test and its hand-built event calendar.

On this page

Terms in this report

null result expected points added

Source

backtests/kyler_cod/FINDINGS.md
updated 2026-10-27 21:26